Tag Archives: amplitudes

Amplitudes 2026, Part II

2 Replies

This is a continuation of my conference coverage from last week. The same warnings apply: this is much more technical than my usual posts, readers beware!

Last week I covered the talks from Monday through Wednesday, so I’ll jump right in here with Thursday morning, where the first speaker was Francesco Riva, who after a brief ad for his new board game Tutti Quantum gave a review of positivity constraints, first-principles restrictions on quantum field theories based on their behavior at high energies. After covering some of the research program’s successes like arguments against Galileons and massive gravity, he talked about how new methods allow one to take into account the possibility of loops of massless particles, essentially by invoking a formal version of the idea that real experiments have finite size. He was followed by Grant Remmen, who talked about his work deriving string theory-like amplitudes from increasingly minimal assumptions. While one can always quibble with the assumptions they impose, I do find it encouraging that they are now managing to do this game with both gravity and gauge theory, and with five-particle amplitudes, not just four.

Paul Heslop then covered applications of bootstrap techniques to non-supersymmetric theories, where analytic superspace still finds a way to be useful. Tomasz Taylor covered progress in calculating Yang-Mills amplitudes in de Sitter space. Andrea Puhm and Nima Arkani-Hamed don’t have slides online yet: Puhm’s title suggests her talk was part of the celestial holography field, while Nima’s was likely similar to his talk at Lancefest, where he talked about calculations of amplitudes in the limit of a very large number of loops (represented in the field with a capital L) or very large numbers of particles (represented in the field with a lower-case n). Given the different context, I’m guessing he left out the self-effacing jokes where he was “little n” and Lance was “big L”, though I’m hoping he at least mentioned his students had been checking their results against what they called the “lanswer”.

The evening ended with a gong show, which for the non-initiated is a series of short student talks with a strict time limit (hence the gong). I’m not quite so intrepid as to read all of the slides for these, commenters who attended are welcome to highlight special examples.

Friday began with a talk by Agnese Bissi, who reviewed the connection between holographic correlators and amplitudes in AdS space. I liked her emphasis on this as a lab to find nice new representations of amplitudes, and her summary at the end of the current frontiers. Axel Kleinschmidt followed with a talk on one-loop string amplitudes, where he and his collaborators have gotten gradually more proficient at manipulating the rich structure of elliptic functions that make an appearance. Piotr Tourkine talked about his work using the S-matrix bootstrap to find scattering amplitudes in higher dimensions, a context where accounting for thresholds presented a new challenge. This is a problem people are approaching with genuine supercomputers, he quoted one calculation at 100,000 CPU hours. Lauren Williams is one of a small community of mathematicians who have been intrigued by the amplituhedron, her talk was a walk through a series of conjectures, some made by physicists, some by mathematicians, most with counterexamples found in the last few years.

Finally, Zvi Bern closed the conference with a talk on the frontier of amplitudes-based gravitational wave calculations, referring to it, probably to David Kosower’s annoyance, as 5PM. Some parts of this frontier have been calculated, but a few have proved hard going, bottlenecked by immensely challenging integrals, which go beyond the capabilities of publicly available codes. The new integrals have a variety of strange functions, including the elliptics and Calabi-Yaus I spent time on in my own career, as well as Heun integrals, which I imagine I will have to learn more about from the slides of Elliptics & beyond ’26, as I don’t remember people talking about them when I was in the field. New integration strategies have led to improvements in the public codes and seem to be making good progress, but Zvi highlighted that ideally they want to not need supercomputers at all, as they’ll need to go higher in loops to see effects from, for example, the deformability of neutron stars.

Amplitudes 2027 will be in Munich, with the summer school in Mainz under Stefan Weinzierl’s capable hands. I’m looking forward to seeing what the state of the art looks like then!

Amplitudes 2026

2 Replies

This week was Amplitudes, my old subfield’s big yearly conference. This year, it’s at Queen Mary University of London.

I’m too busy to attend these days, now that nobody is paying me a salary to do that sort of thing. But it’s still a good chance to keep up with the field, which helps me find stories. And I know I have a few readers who are interested. So I read the slides when I can, and fill you all in.

As of writing this, I’ve read through slides from the first few days of talks. I’ll likely post on the rest next week. As usual when I’m conference-blogging, this post will be quite a bit more technical than my average post, so readers beware: I’ll be mentioning a lot of amplitude-ish ideas without much explanation. I’m happy to explain in the comments if you’re curious, though!

Before I launch into talking about the content, I should mention something I’ve heard about the venue. Apparently this year registration filled up surprisingly quickly. The rumor is that the folks at Queen Mary weren’t able to find a venue with enough space to host the full community, leading to a smaller conference than usual. As the Amplitudes subfield grows, I suspect this will be more and more of a challenge. I remember back when I was organizing in 2021, rooms that could hold enough people were shockingly expensive to book. I wouldn’t be surprised if this becomes more of an issue going forward.

David Kosower opened the conference with a review of the state of the art in amplitudes for gravitational waves. I enjoyed an early slide showing how LIGO has increased in precision over the last ten years, which really helped illustrate the value that good theoretical predictions can bring as the experiment gets better. I also appreciated his attempt to get people to stop saying “post-Minkowskian” and start using a more normal term like Relativistic Perturbation Theory, though based on the other talks it doesn’t look like it’s catching on. After covering the overall state of the art (currently around four loops) he talked about his own work looking for ways to more directly get waveforms for orbiting black holes out of an amplitudes-style calculation, a theme his collaborator Donal O’Connell covered in more detail later that day. (The trick, apparently, is background fields!) Other gravitational wave-related talks came from Gustav Jakobsen, who covered four-loop results using the worldline method, Graham Brown, who explained how to calculate the Magnusian Jakobsen mentioned (it’s the log of the amplitude, basically), Canxin Shi, who talked about how a concept called Stratonovich-Weyl quantization helps explain structures that keep showing up in classical observables, Giulia Isabella, who covered a way to get six-loop contributions to gravitational waves via a wave equation, Lara Bohnenblust, who showed how to get strong-field information from a shockwave limit, and Gang Chen, whose slides were brief enough that I didn’t really get a clear idea of what he was up to. Rounding out the day were two talks that didn’t involve gravitational waves: Zihan Zhou, reporting on work with Nima on a water wave polytope called the hydrotope that they found a general formula for with help from Claude (the AI, not Duhr, to steal a recurring joke from Lancefest), and Michael Ruf, who probably snuck in due to the fact that he has done a lot of work with gravitational waves, but was reporting on a QCD calculation using a tool called Scorpio.

Tuesday had more of a QCD theme. Thomas Gehrmann gave the day’s opening review talk, where he pointed out that to get 1% precision for predictions for the LHC, we’ll likely need three-loop calculations. He pointed out the important role IR real emission calculations and improvements in parton distribution functions need to play, and pointed out that collider physics isn’t just the LHC: between reanalyses of older electron-positron collider data and the upcoming electron-ion collider and FCC, there are even more contexts where high-precision QCD will matter. Simone Zoia talked about one aspect of the current state of the art, five-particle processes involving massive particles, where the functions get strange and elliptical and one has to think hard about the methods one uses (for example, do you want canonical differential equations, or almost-canonical? Bulirsch-Stoer or AMFlow for numerics?) He also included an excellent Laplace quote, “Nature Laughs at the Difficulties of Integration”. Yang Zhang talked about a different calculational frontier, covering progress in the planar limit and with no masses, but for two-loop six-particle and three-loop five-particle scattering. The talk included some nice branding, like an “epsilon collaboration” proposing an algorithm to find epsilon forms for any Feynman integral and a program called Effortless to find symbol letters. Yang Zhang is also a skilled amateur photographer, and he mentions taking the conference photo for Amplitudes five times before. Personally, I’m surprised it’s only been five times! Dmitry Chicherin presented results bootstrapping QCD amplitudes. This was a dream of mine back in my hexagon function days, and while they aren’t quite at the point of being useful (my understanding is they’re still only getting the leading transcendentality, and none of the amplitudes they’re finding are new) it’s still pretty cool that this is even possible now. Bo Feng proposed what he claims is a general algorithm to find generating functions for IBP reduction. It’s not clear to me whether his setup bypasses the computational difficulties in existing methods (mostly involving solving large systems of equations), or whether it shifts the issue elsewhere, though. Pierre Vanhove’s talk also involved QCD, though at an effective field theory mediated step removed, with chiral perturbation theory, a theory of low-energy QCD that he used to shore up the lower energy ranges of lattice calculations for contributions to the muon anomalous magnetic moment (a natural calculation to involve him due to the presence of novel elliptic integrals). Overall, the QCD talks impressed me with the wide range of software packages mentioned, some of which already existed when I was in the field but many of which are new. I do wonder if this is just a result of many research paths maturing at the same time, or if people are finding it easier to write packages with tools like Claude Code.

The remaining three Tuesday talks were more mathematical or theoretical in focus. Henrik Johansson talked about black hole Compton scattering in N=8 supergravity, where it’s possible to use the wave equation to extract all-loop results. Cristian Vergu reported on his progress with Landau analysis. It’s been really fun watching this grow from a small reading group at NBI to what appears to be at this point quite a deep understanding, including a picture of how to understand amplitude singularities with quite a lot of breadth and detail, and the persistent hope that this could allow one to manipulate singular quantities without needing dimensional regularization. Michael Borinsky gave an update on tropicalization, where he showcased a theorem he proved demonstrating a way to compute certain classes of amplitudes in polynomial time in the loop order, even when the number of Feynman diagrams increases factorially. The examples he talked through were impressive, but I’m still a bit skeptical this could work for the Standard Model. I’d want to talk to him about it to figure it out, anyway!

Wednesday was a short day, as is tradition, to give people some time for tourism in the afternoon. Alexander Zhiboedov began the day with a talk on energy correlators in N=4 super Yang-Mills at finite coupling, where he is now able to do a bootstrap with two-sided bounds to hone in on the actual quantity even outside of the planar limit. Arthur Lipstein and Daniel Baumann both talked about cosmological correlators, the former with amplitudeologists’ favorite toy model of the conformally coupled scalar and the latter with Yang-Mills and gravity in de Sitter space. Dave Dunbar closed the day with a historical talk, walking through the milestone amplitudes papers before the Amplitudes conference existed. It’s the kind of talk he could have given at Lancefest the week before if others hadn’t already covered the material!

I’ll cover the rest of the conference (Thursday and Friday) in next week’s post, so see you then!

At Lancefest

Leave a reply

It’s been a while since I’ve said this: I’m at a conference this week!

Specifically, I’m at Lancefest, which is not just any old conference, but a birthday conference for Lance Dixon. When a renowned academic turns 60 or so, their students and collaborators hold a birthday party-flavored conference for them. The conferences are usually a mix of academic talks and reminiscences, with the occasional roast thrown in.

I went to my advisor’s birthday conference four years ago. Lance wasn’t my advisor, but in many ways he might as well have been. When my advisor took a sabbatical in the middle of my PhD, he sent me to work with Lance. It was my first real experience doing research in a team, not just puzzling away by myself with occasional feedback. And I was hooked: I spent the rest of my academic career in Lance’s field. We collaborated time after time, and even when I started to branch out he remained a frequent presence.

In part, that’s because Lance’s field was really Lance’s field. Amplitudeology has grown a lot since I started out, with several hundred people going to the field’s big yearly conference and subfields like Elliptics having their own yearly conferences. In such a world, it’s tough for anyone to feel like a truly central figure. But Lance tends to. He’s been able to keep up with that growing world, to keep finding important problems and keep understanding others’ ideas. While some have specialized, or stepped back, Lance seems to somehow manage to be a father figure for the whole field all at once. It’s a capability his advisor Jeff Harvey might have predicted, he mentioned in his talk that Lance’s interests were always broad. Over the years the young students I met who joined when the amplitudes field was already large saw Lance as a kind of mysterious titan, and were occasionally awed that I had worked with him. “What was that like?”

Well, it was like working with Lance. Lance isn’t a manager at heart, like some senior academics end up. He wants to understand everything he works on, and will happily dive in, Maple subscription in hand, and try to figure things out for himself. He wants you to keep up with him, understanding on your own terms, to keep him honest, to provide an independent check. But that often wasn’t possible, because the man is just so damn fast. He’d be miles ahead of me, with his Maple and his laptop, while I was churning overnight Mathematica runs on a twenty-machine cluster.

(Was the difference due to our taste in software? Partly. I did learn to use Maple later, it genuinely is faster at some things. But Lance is faster at almost everything.)

And he cared so damn much sometimes. About getting things right, like a scientist should. About making things nice, too: finding a pretty basis of functions, a better notation for the paper, something that might jostle out the next big insight. Working with him, you could feel like that one paper was the most important thing in the universe.

Others at the event have had similar stories. Fernando Febres Cordero remembers noticing a potential issue, emailing Lance about it, and in a few minutes hearing back with a potential explanation.

Lance is someone who became a leader without really being a politician. He doesn’t have the legions of students in tenured positions that some do. I trimmed that count by one, and it wasn’t huge to begin with. But for someone who isn’t “everywhere” in that sense, he manages to be “everywhere” all the same.

So Lance, happy birthday! You’re really the only person who could have had a birthday conference quite like this, a cross-section of the field, all with something kind to say. Thanks for putting up with any embarrassment associated with having this much attention for three days, and I wish you many Maple-fueled mysteries to come.

Breakthrough Prize 2026

1 Reply

Because of last week’s “bonus info” post, I’m only now getting around to commenting on this year’s Breakthrough Prizes in Fundamental Physics. While I don’t comment on them every year, I know enough about several of this year’s winners that I figured a post would be helpful.

For those who haven’t heard of it, the Breakthrough Prizes are a bit like the Nobel, if it was created by a 21st century rich person instead of a 19th century one. They give out more money, and instead of an organization like the Swedish Academy of Sciences they pick winners via a committee of past winners. They’re more flexible in structure than the Nobel, with extra prizes for early-career researchers and a tendency to reward accomplishments that are either entirely theoretical or solid experimental work that doesn’t show a new discovery, both of which are things the Nobel Prize is structured to avoid. They’ve also shown willingness to reward large collaborations, rather than following the Nobel’s informal rule to only give the award to three people at a time.

This last was on display this year in their main award in physics this year, for the muon g-2 collaborations. The award is going to collaborations of scientists and engineers at three different particle colliders, for work done over a span of over fifty years to measure the magnetic properties of the muon. These measurements have shown a tantalizing discrepancy with predictions that inspired many to conjecture new physics. However, in the last few years it’s looked more and more like the discrepancy was due to an imprecise prediction, and better methods seem to be converging to the experimental value. At this point, smart money is that there is no disagreement with the Standard Model here, but as always in science there’s a chance some mystery remains.

The Breakthrough Prize also offered a special, out-of-schedule prize to David Gross. Already a Nobel laureate, Gross had a crucial role in our understanding of the force of quantum chromodynamics that binds protons and neutrons together. He was also a major founding figure in string theory, and since the Breakthrough Prize is more comfortable recognizing theoretical contributions they get to mention this as well. Gross is also known in the community for his personality, which tends to fill up any room he’s in. I can only imagine the conversations that led to Breakthrough’s decision to add a special prize for him this year.

Breakthrough is also adding a new recurring prize, the Vera Rubin New Frontiers Prize, honoring women who make important contributions to physics within two years of their PhD. The prize is a bit smaller than the exiting early-career New Horizons in Physics Prizes, presumably because it goes to even younger researchers. This year’s winner is from my old field, scattering amplitudes. Carolina Figueiredo is part of the latest evolution of the research program behind the amplituhedron. The new framework of “surfaceology” seems like a promising geometry-flavored way to understand particle physics calculations in more realistic theories, and unlike its predecessors may have some practical value eventually as well. Congrats Carolina!

Finally, the New Horizons in Physics Prizes are for impressive early-career researchers. I don’t know much about the first recipient, Benjamin Safdi, who works on searches for axions and axion-like particles, today’s most trendy dark matter candidate. I know a bit more about the work done by Clay Córdova, Thomas Dumitrescu, Shu-Heng Shao, and Yifan Wang, having met several of them in my physics career. They work on what are called generalized symmetries, concepts which go beyond the usual idea of how symmetry is supposed to work by involving more complicated tensors. I saw these crop up a fair bit in talks, but they were distant enough from my area that I never had a particularly clear grasp of what people were doing with them. I know even less about the work of the last three, Dillon Brout, J. Colin Hill, Mathew Madhavacheril, Maria Vincenzi, Daniel Scolnic, and W. L. Kimmy Wu, on cosmological measurements, but I was friends with Mathew in grad school and am impressed that he’s now working on cosmology given how little cosmology research there was at Stony Brook at the time.

About the OpenAI Amplitudes Paper, but Not as Much as You’d Like

5 Replies

I’ve had a bit more time to dig in to the paper I mentioned last week, where OpenAI collaborated with amplitudes researchers, using one of their internal models to find and prove a simplified version of a particle physics formula. I figured I’d say a bit about my own impressions from reading the paper and OpenAI’s press release.

This won’t be a real “deep dive”, though it will be long nonetheless. As it turns out, most of the questions I’d like answers to aren’t answered in the paper or the press release. Getting them will involve actual journalistic work, i.e. blocking off time to interview people, and I haven’t done that yet. What I can do is talk about what I know so far, and what I’m still wondering.

Context:

Scattering amplitudes are formulas used by particle physicists to make predictions. For a while, people would just calculate these when they needed them, writing down pages of mess that you could plug in numbers to to get answers. However, forty years ago two physicists decided they wanted more, writing “we hope to obtain a simplified form for the answer, making our result not only an experimentalist’s, but a theorist’s delight.”

In their next paper, they managed to find that “theorist’s delight”: a simplified, intuitive-looking answer that worked for calculations involving any number of particles, summarizing many different calculations. Ten years later, a few people had started building on it, and ten years after that, the big shots started paying attention. A whole subfield, “amplitudeology”, grew from that seed, finding new forms of “theorists’s delight” in scattering amplitudes.

Each subfield has its own kind of “theory of victory”, its own concept for what kind of research is most likely to yield progress. In amplitudes, it’s these kinds of simplifications. When they work out well, they yield new, more efficient calculation techniques, yielding new messy results which can be simplified once more. To one extent or another, most of the field is chasing after those situations when simplification works out well.

That motivation shapes both the most ambitious projects of senior researchers, and the smallest student projects. Students often spend enormous amounts of time looking for a nice formula for something and figuring out how to generalize it, often on a question suggested by a senior researcher. These projects mostly serve as training, but occasionally manage to uncover something more impressive and useful, an idea others can build around.

I’m mentioning all of this, because as far as I can tell, what ChatGPT and the OpenAI internal model contributed here roughly lines up with the roles students have on amplitudes papers. In fact, it’s not that different from the role one of the authors, Alfredo Guevara, had when I helped mentor him during his Master’s.

Senior researchers noticed something unusual, suggested by prior literature. They decided to work out the implications, did some calculations, and got some messy results. It wasn’t immediately clear how to clean up the results, or generalize them. So they waited, and eventually were contacted by someone eager for a research project, who did the work to get the results into a nice, general form. Then everyone publishes together on a shared paper.

How impressed should you be?

I said, “as far as I can tell” above. What’s annoying is that this paper makes it hard to tell.

If you read through the paper, they mention AI briefly in the introduction, saying they used GPT-5.2 Pro to conjecture formula (39) in the paper, and an OpenAI internal model to prove it. The press release actually goes into more detail, saying that the humans found formulas (29)-(32), and GPT-5.2 Pro found a special case where it could simplify them to formulas (35)-(38), before conjecturing (39). You can get even more detail from an X thread by one of the authors, OpenAI Research Scientist Alex Lupsasca. Alex had done his PhD with another one of the authors, Andrew Strominger, and was excited to apply the tools he was developing at OpenAI to his old research field. So they looked for a problem, and tried out the one that ended up in the paper.

What is missing, from the paper, press release, and X thread, is any real detail about how the AI tools were used. We don’t have the prompts, or the output, or any real way to assess how much input came from humans and how much from the AI.

(We have more for their follow-up paper, where Lupsasca posted a transcript of the chat.)

Contra some commentators, I don’t think the authors are being intentionally vague here. They’re following business as usual. In a theoretical physics paper, you don’t list who did what, or take detailed account of how you came to the results. You clean things up, and create a nice narrative. This goes double if you’re aiming for one of the most prestigious journals, which tend to have length limits.

This business-as-usual approach is ok, if frustrating, for the average physics paper. It is, however, entirely inappropriate for a paper showcasing emerging technologies. For a paper that was going to be highlighted this highly by OpenAI, the question of how they reached their conclusion is much more interesting than the results themselves. And while I wouldn’t ask them to go to the standards of an actual AI paper, with ablation analysis and all that jazz, they could at least have aimed for the level of detail of my final research paper, which gave samples of the AI input and output used in its genetic algorithm.

For the moment, then, I have to guess what input the AI had, and what it actually accomplished.

Let’s focus on the work done by the internal OpenAI model. The descriptions I’ve seen suggest that it started where GPT-5.2 Pro did, with formulas (29)-(32), but with a more specific prompt that guided what it was looking for. It then ran for 12 hours with no additional input, and both conjectured (39) and proved it was correct, providing essentially the proof that follows formula (39) in the paper.

Given that, how impressed should we be?

First, the model needs to decide to go to a specialized region, instead of trying to simplify the formula in full generality. I don’t know whether they prompted their internal model explicitly to do this. It’s not something I’d expect a student to do, because students don’t know what types of results are interesting enough to get published, so they wouldn’t be confident in computing only a limited version of a result without an advisor telling them it was ok. On the other hand, it is actually something I’d expect an LLM to be unusually likely to do, as a result of not managing to consistently stick to the original request! What I don’t know is whether the LLM proposed this for the right reason: that if you have the formula for one region, you can usually find it for other regions.

Second, the model needs to take formulas (29)-(32), write them in the specialized region, and simplify them to formulas (35)-(38). I’ve seen a few people saying you can do this pretty easily with Mathematica. That’s true, though not every senior researcher is comfortable doing that kind of thing, as you need to be a bit smarter than just using the Simplify[] command. Most of the people on this paper strike me as pen-and-paper types who wouldn’t necessarily know how to do that. It’s definitely the kind of thing I’d expect most students to figure out, perhaps after a couple of weeks of flailing around if it’s their first crack at it. The LLM likely would not have used Mathematica, but would have used SymPy, since these “AI scientist” setups usually can write and execute Python code. You shouldn’t think of this as the AI reasoning through the calculation itself, but it at least sounds like it was reasonably quick at coding it up.

Then, the model needs to conjecture formula (39). This gets highlighted in the intro, but as many have pointed out, it’s pretty easy to do. If any non-physicists are still reading at this point, take a look:

Could you guess (39) from (35)-(38)?

After that, the paper goes over the proof that formula (39) is correct. Most of this proof isn’t terribly difficult, but the way it begins is actually unusual in an interesting way. The proof uses ideas from time-ordered perturbation theory, an old-fashioned way to do particle physics calculations. Time-ordered perturbation theory isn’t something any of the authors are known for using with regularity, but it has recently seen a resurgence in another area of amplitudes research, showing up for example in papers by Matthew Schwartz, a colleague of Strominger at Harvard.

If a student of Strominger came up with an idea drawn from time-ordered perturbation theory, that would actually be pretty impressive. It would mean that, rather than just learning from their official mentor, this student was talking to other people in the department and broadening their horizons, showing a kind of initiative that theoretical physicists value a lot.

From an LLM, though, this is not impressive in the same way. The LLM was not trained by Strominger, it did not learn specifically from Strominger’s papers. Its context suggested it was working on an amplitudes paper, and it produced an idea which would be at home in an amplitudes paper, just a different one than the one it was working on.

While not impressive, that capability may be quite useful. Academic subfields can often get very specialized and siloed. A tool that suggests ideas from elsewhere in the field could help some people broaden their horizons.

Overall, it appears that that twelve-hour OpenAI internal model run reproduced roughly what an unusually bright student would be able to contribute over the course of a several-month project. Like most student projects, you could find a senior researcher who could do the project much faster, maybe even faster than the LLM. But it’s unclear whether any of the authors could have: different senior researchers have different skillsets.

A stab at implications:

If we take all this at face-value, it looks like OpenAI’s internal model was able to do a reasonably competent student project with no serious mistakes in twelve hours. If they started selling that capability, what would happen?

If it’s cheap enough, you might wonder if professors would choose to use the OpenAI model instead of hiring students. I don’t think this would happen, though: I think it misunderstands why these kinds of student projects exist in a theoretical field. Professors sometimes use students to get results they care about, but more often, the student’s interest is itself the motivation, with the professor wanting to educate someone, to empire-build, or just to take on their share of the department’s responsibilities. AI is only useful for this insofar as AI companies continue reaching out to these people to generate press releases: once this is routinely possible, the motivation goes away.

More dangerously, if it’s even cheaper, you could imagine students being tempted to use it. The whole point of a student project is to train and acculturate the student, to get them to the point where they have affection for the field and the capability to do more impressive things. You can’t skip that, but people are going to be tempted to.

And of course, there is the broader question of how much farther this technology can go. That’s the hardest to estimate here, since we don’t know the prompts used. So I don’t know if seeing this result tells us anything more about the bigger picture than we knew going in.

Remaining questions:

At the end of the day, there are a lot of things I still want to know. And if I do end up covering this professionally, they’re things I’ll ask.

  1. What was the prompt given to the internal model, and how much did it do based on that prompt?
  2. Was it really done in one shot, no retries or feedback?
  3. How much did running the internal model cost?
  4. Is this result likely to be useful? Are there things people want to calculate that this could make easier? Recursion relations it could seed? Is it useful for SCET somehow?
  5. How easy would it have been for the authors to do what the LLM did? What about other experts in the community?

Hypothesis: If AI Is Bad at Originality, It’s a Documentation Problem

2 Replies

Recently, a few people have asked me about this paper.

A couple weeks back, OpenAI announced a collaboration with a group of amplitudes researchers, physicists who study the types of calculations people do to make predictions at particle colliders. The amplitudes folks had identified an interesting loophole, finding a calculation that many would have expected to be zero actually gave a nonzero answer. They did the calculation for different examples involving more and more particles, and got some fairly messy answers. They suspected, as amplitudes researchers always expect, that there was a simpler formula, one that worked for any number of particles. But they couldn’t find it.

Then a former amplitudes researcher at OpenAI suggested that they use AI to find it.

“Use AI” can mean a lot of different things, and most of them don’t look much like the way the average person talks to ChatGPT. This was closer than most. They were using “reasoning models”, loops that try to predict the next few phrases in a “chain of thought” again and again and again. Using that kind of tool, they were able to find that simpler formula, and mathematically prove that it was correct.

A few of you are hoping for an in-depth post about what they did, and its implications. This isn’t that. I’m still figuring out if I’ll be writing that for an actual news site, for money, rather than free, for you folks.

Instead, I want to talk about a specific idea I’ve seen crop up around the paper.

See, for some, the existence of a result like this isn’t all that surprising.

Mathematicians have been experimenting with reasoning models for a bit, now. Recently, a group published a systematic study, setting the AI loose on a database of minor open problems proposed by the famously amphetamine-fueled mathematician Paul Erdös. The AI managed to tackle a few of the problems, sometimes by identifying existing solutions that had not yet been linked to the problem database, but sometimes by proofs that appeared to be new.

The Erdös problems solved by the AI were not especially important. Neither was the problem solved by the amplitudes researchers, as far as I can tell at this point.

But I get the impression the amplitudes problem was a bit more interesting than the Erdös problems. The difference, so far, has mostly been attributed to human involvement. This amplitudes paper started because human amplitudes researchers found an interesting loophole, and only after that used the AI. Unlike the mathematicians, they weren’t just searching a database.

This lines up with a general point, one people tend to make much less carefully. It’s often said that, unlike humans, AI will never be truly creative. It can solve mechanical problems, do things people have done before, but it will never be good at having truly novel ideas.

To me, that line of thinking goes a bit too far. I suspect it’s right on one level, that it will be hard for any of these reasoning models to propose anything truly novel. But if so, I think it will be for a different reason.

The thing is, creativity is not as magical as we make it out to be. Our ideas, scientific or artistic, don’t just come from the gods. They recombine existing ideas, shuffling them in ways more akin to randomness than miracle. They’re then filtered through experience, deep heuristics honed over careers. Some people are good at ideas, and some are bad at them. Having ideas takes work, and there are things people do to improve their ideas. Nothing about creativity suggests it should be impossible to mechanize.

However, a machine trained on text won’t necessarily know how to do any of that.

That’s because in science, we don’t write down our inspirations. By the time a result gets into a scientific paper or textbook, it’s polished and refined into a pure argument, cutting out most of the twists and turns that were an essential part of the creative process. Mathematics is even worse, most math papers don’t even mention the motivation behind the work, let alone the path taken to the paper.

This lack of documentation makes it hard for students, making success much more a function of having the right mentors to model good practices, rather than being able to pick them up from literature everyone can access. I suspect it makes it even harder for language models. And if today’s language model-based reasoning tools are bad at that crucial, human-seeming step, of coming up with the right idea at the right time? I think that has more to do with this lack of documentation, than with the fact that they’re “statistical parrots”.

Amplitudes 2025 This Week

1 Reply

Summer is conference season for academics, and this week held my old sub-field’s big yearly conference, called Amplitudes. This year, it was in Seoul at Seoul National University, the first time the conference has been in Asia.

(I wasn’t there, I don’t go to these anymore. But I’ve been skimming slides in my free time, to give you folks the updates you crave. Be forewarned that conference posts like these get technical fast, I’ll be back to my usual accessible self next week.)

There isn’t a huge amplitudes community in Korea, but it’s bigger than it was back when I got started in the field. Of the organizers, Kanghoon Lee of the Asia Pacific Center for Theoretical Physics and Sangmin Lee of Seoul National University have what I think of as “core amplitudes interests”, like recursion relations and the double-copy. The other Korean organizers are from adjacent areas, work that overlaps with amplitudes but doesn’t show up at the conference each year. There was also a sizeable group of organizers from Taiwan, where there has been a significant amplitudes presence for some time now. I do wonder if Korea was chosen as a compromise between a conference hosted in Taiwan or in mainland China, where there is also quite a substantial amplitudes community.

One thing that impresses me every year is how big, and how sophisticated, the gravitational-wave community in amplitudes has grown. Federico Buccioni’s talk began with a plot that illustrates this well (though that wasn’t his goal):

At the conference Amplitudes, dedicated to the topic of scattering amplitudes, there were almost as many talks with the phrase “black hole” in the title as there were with “scattering” or “amplitudes”! This is for a topic that did not even exist in the subfield when I got my PhD eleven years ago.

With that said, gravitational wave astronomy wasn’t quite as dominant at the conference as Buccioni’s bar chart suggests. There were a few talks each day on the topic: I counted seven in total, excluding any short talks on the subject in the gong show. Spinning black holes were a significant focus, central to Jung-Wook Kim’s, Andres Luna’s and Mao Zeng’s talks (the latter two showing some interesting links between the amplitudes story and classic ideas in classical mechanics) and relevant in several others, with Riccardo Gonzo, Miguel Correia, Ira Rothstein, and Enrico Herrmann’s talks showing not just a wide range of approaches, but an increasing depth of research in this area.

Herrmann’s talk in particular dealt with detector event shapes, a framework that lets physicists think more directly about what a specific particle detector or observer can see. He applied the idea not just to gravitational waves but to quantum gravity and collider physics as well. The latter is historically where this idea has been applied the most thoroughly, as highlighted in Hua Xing Zhu’s talk, where he used them to pick out particular phenomena of interest in QCD.

QCD is, of course, always of interest in the amplitudes field. Buccioni’s talk dealt with the theory’s behavior at high-energies, with a nice example of the “maximal transcendentality principle” where some quantities in QCD are identical to quantities in N=4 super Yang-Mills in the “most transcendental” pieces (loosely, those with the highest powers of pi). Andrea Guerreri’s talk also dealt with high-energy behavior in QCD, trying to address an experimental puzzle where QCD results appeared to violate a fundamental bound all sensible theories were expected to obey. By using S-matrix bootstrap techniques, they clarify the nature of the bound, finding that QCD still obeys it once correctly understood, and conjecture a weird theory that should be possible to frame right on the edge of the bound. The S-matrix bootstrap was also used by Alexandre Homrich, who talked about getting the framework to work for multi-particle scattering.

Heribertus Bayu Hartanto is another recent addition to Korea’s amplitudes community. He talked about a concrete calculation, two-loop five-particle scattering including top quarks, a tricky case that includes elliptic curves.

When amplitudes lead to integrals involving elliptic curves, many standard methods fail. Jake Bourjaily’s talk raised a question he has brought up again and again: what does it mean to do an integral for a new type of function? One possible answer is that it depends on what kind of numerics you can do, and since more general numerical methods can be cumbersome one often needs to understand the new type of function in more detail. In light of that, Stephen Jones’ talk was interesting in taking a common problem often cited with generic approaches (that they have trouble with the complex numbers introduced by Minkowski space) and finding a more natural way in a particular generic approach (sector decomposition) to take them into account. Giulio Salvatori talked about a much less conventional numerical method, linked to the latest trend in Nima-ology, surfaceology. One of the big selling points of the surface integral framework promoted by people like Salvatori and Nima Arkani-Hamed is that it’s supposed to give a clear integral to do for each scattering amplitude, one which should be amenable to a numerical treatment recently developed by Michael Borinsky. Salvatori can currently apply the method only to a toy model (up to ten loops!), but he has some ideas for how to generalize it, which will require handling divergences and numerators.

Other approaches to the “problem of integration” included Anna-Laura Sattelberger’s talk that presented a method to find differential equations for the kind of integrals that show up in amplitudes using the mathematical software Macaulay2, including presenting a package. Matthias Wilhelm talked about the work I did with him, using machine learning to find better methods for solving integrals with integration-by-parts, an area where two other groups have now also published. Pierpaolo Mastrolia talked about integration-by-parts’ up-and-coming contender, intersection theory, a method which appears to be delving into more mathematical tools in an effort to catch up with its competitor.

Sometimes, one is more specifically interested in the singularities of integrals than their numerics more generally. Felix Tellander talked about a geometric method to pin these down which largely went over my head, but he did have a very nice short description of the approach: “Describe the singularities of the integrand. Find a map representing integration. Map the singularities of the integrand onto the singularities of the integral.”

While QCD and gravity are the applications of choice, amplitudes methods germinate in N=4 super Yang-Mills. Ruth Britto’s talk opened the conference with an overview of progress along those lines before going into her own recent work with one-loop integrals and interesting implications of ideas from cluster algebras. Cluster algebras made appearances in several other talks, including Anastasia Volovich’s talk which discussed how ideas from that corner called flag cluster algebras may give insights into QCD amplitudes, though some symbol letters still seem to be hard to track down. Matteo Parisi covered another idea, cluster promotion maps, which he thinks may help pin down algebraic symbol letters.

The link between cluster algebras and symbol letters is an ongoing mystery where the field is seeing progress. Another symbol letter mystery is antipodal duality, where flipping an amplitude like a palindrome somehow gives another valid amplitude. Lance Dixon has made progress in understanding where this duality comes from, finding a toy model where it can be understood and proved.

Others pushed the boundaries of methods specific to N=4 super Yang-Mills, looking for novel structures. Song He’s talk pushes an older approach by Bourjaily and collaborators up to twelve loops, finding new patterns and connections to other theories and observables. Qinglin Yang bootstraps Wilson loops with a Lagrangian insertion, adding a side to the polygon used in previous efforts and finding that, much like when you add particles to amplitudes in a bootstrap, the method gets stricter and more powerful. Jaroslav Trnka talked about work he has been doing with “negative geometries”, an odd method descended from the amplituhedron that looks at amplitudes from a totally different perspective, probing a bit of their non-perturbative data. He’s finding more parts of that setup that can be accessed and re-summed, finding interestingly that multiple-zeta-values show up in quantities where we know they ultimately cancel out. Livia Ferro also talked about a descendant of the amplituhedron, this time for cosmology, getting differential equations for cosmological observables in a particular theory from a combinatorial approach.

Outside of everybody’s favorite theories, some speakers talked about more general approaches to understanding the differences between theories. Andreas Helset covered work on the geometry of the space of quantum fields in a theory, applying the method to a general framework for characterizing deviations from the standard model called the SMEFT. Jasper Roosmale Nepveu also talked about a general space of theories, thinking about how positivity (a trait linked to fundamental constraints like causality and unitarity) gets tangled up with loop effects, and the implications this has for renormalization.

Soft theorems, universal behavior of amplitudes when a particle has low energy, continue to be a trendy topic, with Silvia Nagy showing how the story continues to higher orders and Sangmin Choi investigating loop effects. Callum Jones talks about one of the more powerful results from the soft limit, Weinberg’s theorem showing the uniqueness of gravity. Weinberg’s proof was set up in Minkowski space, but we may ultimately live in curved, de Sitter space. Jones showed how the ideas Weinberg explored generalize in de Sitter, using some tools from the soft-theorem-inspired field of dS/CFT. Julio Parra-Martinez, meanwhile, tied soft theorems to another trendy topic, higher symmetries, a more general notion of the usual types of symmetries that physicists have explored in the past. Lucia Cordova reported work that was not particularly connected to soft theorems but was connected to these higher symmetries, showing how they interact with crossing symmetry and the S-matrix bootstrap.

Finally, a surprisingly large number of talks linked to Kevin Costello and Natalie Paquette’s work with self-dual gauge theories, where they found exact solutions from a fairly mathy angle. Paquette gave an update on her work on the topic, while Alfredo Guevara talked about applications to black holes, comparing the power of expanding around a self-dual gauge theory to that of working with supersymmetry. Atul Sharma looked at scattering in self-dual backgrounds in work that merges older twistor space ideas with the new approach, while Roland Bittelson talked about calculating around an instanton background.

Also, I had another piece up this week at FirstPrinciples, based on an interview with the (outgoing) president of the Sloan Foundation. I won’t have a “bonus info” post for this one, as most of what I learned went into the piece. But if you don’t know what the Sloan Foundation does, take a look! I hadn’t known they funded Jupyter notebooks and Hidden Figures, or that they introduced Kahneman and Tversky.

Integration by Parts, Evolved

5 Replies

I posted what may be my last academic paper today, about a project I’ve been working on with Matthias Wilhelm for most of the last year. The paper is now online here. For me, the project has been a chance to broaden my horizons, learn new skills, and start to step out of my academic comfort zone. For Matthias, I hope it was grant money well spent.

I wanted to work on something related to machine learning, for the usual trendy employability reasons. Matthias was already working with machine learning, but was interested in pursuing a different question.

When is machine learning worthwhile? Machine learning methods are heuristics, unreliable methods that sometimes work well. You don’t use a heuristic if you have a reliable method that runs fast enough. But if all you have are heuristics to begin with, then machine learning can give you a better heuristic.

Matthias noticed a heuristic embedded deep in how we do particle physics, and guessed that we could do better. In particle physics, we use pictures called Feynman diagrams to predict the probabilities for different outcomes of collisions, comparing those predictions to observation to look for evidence of new physics. Each Feynman diagram corresponds to an integral, and for each calculation there are hundreds, thousands, or even millions of those integrals to do.

Luckily, physicists don’t actually have to do all those integrals. It turns out that most of them are related, by a slightly more advanced version of that calculus class mainstay, integration by parts. Using integration by parts you can solve a list of equations, finding out how to write your integrals in terms of a much smaller list.

How big a list of equations do you need, and which ones? Twenty-five years ago, Stefano Laporta proposed a “golden rule” to choose, based on his own experience, and people have been using it (more or less, with their own tweaks) since then.

Laporta’s rule is a heuristic, with no proof that it is the best option, or even that it will always work. So we probably shouldn’t have been surprised when someone came up with a better heuristic. Watching talks at a December 2023 conference, Matthias saw a presentation by Johann Usovitsch on a curious new rule. The rule was surprisingly simple, just one extra condition on top of Laporta’s. But it was enough to reduce the number of equations by a factor of twenty.

That’s great progress, but it’s also a bit frustrating. Over almost twenty-five years, no-one had guessed this one simple change?

Maybe, thought Matthias and I, we need to get better at guessing.

We started out thinking we’d try reinforcement learning, a technique where a machine is trained by playing a game again and again, changing its strategy when that strategy brings it a reward. We thought we could have the machine learn to cut away extra equations, getting rewarded if it could cut more while still getting the right answer. We didn’t end up pursuing this very far before realizing another strategy would be a better fit.

What is a rule, but a program? Laporta’s golden rule and Johann’s new rule could both be expressed as simple programs. So we decided to use a method that could guess programs.

One method stood out for sheer trendiness and audacity: FunSearch. FunSearch is a type of algorithm called a genetic algorithm, which tries to mimic evolution. It makes a population of different programs, “breeds” them with each other to create new programs, and periodically selects out the ones that perform best. That’s not the trendy or audacious part, though, people have been doing that sort of genetic programming for a long time.

The trendy, audacious part is that FunSearch generates these programs with a Large Language Model, or LLM (the type of technology behind ChatGPT). Using an LLM trained to complete code, FunSearch presents the model with two programs labeled v0 and v1 and asks it to complete v2. In general, program v2 will have some traits from v0 and v1, but also a lot of variation due to the unpredictable output of LLMs. The inventors of FunSearch used this to contribute the variation needed for evolution, using it to evolve programs to find better solutions to math problems.

We decided to try FunSearch on our problem, modifying it a bit to fit the case. We asked it to find a shorter list of equations, giving a better score for a shorter list but a penalty if the list wasn’t able to solve the problem fully.

Some tinkering and headaches later, it worked! After a few days and thousands of program guesses, FunSearch was able to find a program that reproduced the new rule Johann had presented. A few hours more, and it even found a rule that was slightly better!

But then we started wondering: do we actually need days of GPU time to do this?

An expert on heuristics we knew had insisted, at the beginning, that we try something simpler. The approach we tried then didn’t work. But after running into some people using genetic programming at a conference last year, we decided to try again, using a Python package they used in their work. This time, it worked like a charm, taking hours rather than days to find good rules.

This was all pretty cool, a great opportunity for me to cut my teeth on Python programming and its various attendant skills. And it’s been inspiring, with Matthias drawing together more people interested in seeing just how much these kinds of heuristic methods can do there. I should be clear though, that so far I don’t think our result is useful. We did better than the state of the art on an example, but only slightly, and in a way that I’d guess doesn’t generalize. And we needed quite a bit of overhead to do it. Ultimately, while I suspect there’s something useful to find in this direction, it’s going to require more collaboration, both with people using the existing methods who know better what the bottlenecks are, and with experts in these, and other, kinds of heuristics.

So I’m curious to see what the future holds. And for the moment, happy that I got to try this out!

How Small Scales Can Matter for Large Scales

Leave a reply

For a certain type of physicist, nothing matters more than finding the ultimate laws of nature for its tiniest building-blocks, the rules that govern quantum gravity and tell us where the other laws of physics come from. But because they know very little about those laws at this point, they can predict almost nothing about observations on the larger distance scales we can actually measure.

“Almost nothing” isn’t nothing, though. Theoretical physicists don’t know nature’s ultimate laws. But some things about them can be reasonably guessed. The ultimate laws should include a theory of quantum gravity. They should explain at least some of what we see in particle physics now, explaining why different particles have different masses in terms of a simpler theory. And they should “make sense”, respecting cause and effect, the laws of probability, and Einstein’s overall picture of space and time.

All of these are assumptions, of course. Further assumptions are needed to derive any testable consequences from them. But a few communities in theoretical physics are willing to take the plunge, and see what consequences their assumptions have.

First, there’s the Swampland. String theorists posit that the world has extra dimensions, which can be curled up in a variety of ways to hide from view, with different observable consequences depending on how the dimensions are curled up. This list of different observable consequences is referred to as the Landscape of possibilities. Based on that, some string theorists coined the term “Swampland” to represent an area outside the Landscape, containing observations that are incompatible with quantum gravity altogether, and tried to figure out what those observations would be.

In principle, the Swampland includes the work of all the other communities on this list, since a theory of quantum gravity ought to be consistent with other principles as well. In practice, people who use the term focus on consequences of gravity in particular. The earliest such ideas argued from thought experiments with black holes, finding results that seemed to demand that gravity be the weakest force for at least one type of particle. Later researchers would more frequently use string theory as an example, looking at what kinds of constructions people had been able to make in the Landscape to guess what might lie outside of it. They’ve used this to argue that dark energy might be temporary, and to try to figure out what traits new particles might have.

Second, I should mention naturalness. When talking about naturalness, people often use the analogy of a pen balanced on its tip. While possible in principle, it must have been set up almost perfectly, since any small imbalance would cause it to topple, and that perfection demands an explanation. Similarly, in particle physics, things like the mass of the Higgs boson and the strength of dark energy seem to be carefully balanced, so that a small change in how they were set up would lead to a much heavier Higgs boson or much stronger dark energy. The need for an explanation for the Higgs’ careful balance is why many physicists expected the Large Hadron Collider to discover additional new particles.

As I’ve argued before, this kind of argument rests on assumptions about the fundamental laws of physics. It assumes that the fundamental laws explain the mass of the Higgs, not merely by giving it an arbitrary number but by showing how that number comes from a non-arbitrary physical process. It also assumes that we understand well how physical processes like that work, and what kinds of numbers they can give. That’s why I think of naturalness as a type of argument, much like the Swampland, that uses the smallest scales to constrain larger ones.

Third is a host of constraints that usually go together: causality, unitarity, and positivity. Causality comes from cause and effect in a relativistic universe. Because two distant events can appear to happen in different orders depending on how fast you’re going, any way to send signals faster than light is also a way to send signals back in time, causing all of the paradoxes familiar from science fiction. Unitarity comes from quantum mechanics. If quantum calculations are supposed to give the probability of things happening, those probabilities should make sense as probabilities: for example, they should never go above one.

You might guess that almost any theory would satisfy these constraints. But if you extend a theory to the smallest scales, some theories that otherwise seem sensible end up failing this test. Actually linking things up takes other conjectures about the mathematical form theories can have, conjectures that seem more solid than the ones underlying Swampland and naturalness constraints but that still can’t be conclusively proven. If you trust the conjectures, you can derive restrictions, often called positivity constraints when they demand that some set of observations is positive. There has been a renaissance in this kind of research over the last few years, including arguments that certain speculative theories of gravity can’t actually work.

Replacing Space-Time With the Space in Your Eyes

5 Replies

Nima Arkani-Hamed thinks space-time is doomed.

That doesn’t mean he thinks it’s about to be destroyed by a supervillain. Rather, Nima, like many physicists, thinks that space and time are just approximations to a deeper reality. In order to make sense of gravity in a quantum world, seemingly fundamental ideas, like that particles move through particular places at particular times, will probably need to become more flexible.

But while most people who think space-time is doomed research quantum gravity, Nima’s path is different. Nima has been studying scattering amplitudes, formulas used by particle physicists to predict how likely particles are to collide in particular ways. He has been trying to find ways to calculate these scattering amplitudes without referring directly to particles traveling through space and time. In the long run, the hope is that knowing how to do these calculations will help suggest new theories beyond particle physics, theories that can’t be described with space and time at all.

Ten years ago, Nima figured out how to do this in a particular theory, one that doesn’t describe the real world. For that theory he was able to find a new picture of how to calculate scattering amplitudes based on a combinatorical, geometric space with no reference to particles traveling through space-time. He gave this space the catchy name “the amplituhedron“. In the years since, he found a few other “hedra” describing different theories.

Now, he’s got a new approach. The new approach doesn’t have the same kind of catchy name: people sometimes call it surfaceology, or curve integral formalism. Like the amplituhedron, it involves concepts from combinatorics and geometry. It isn’t quite as “pure” as the amplituhedron: it uses a bit more from ordinary particle physics, and while it avoids specific paths in space-time it does care about the shape of those paths. Still, it has one big advantage: unlike the amplituhedron, Nima’s new approach looks like it can work for at least a few of the theories that actually describe the real world.

The amplituhedron was mysterious. Instead of space and time, it described the world in terms of a geometric space whose meaning was unclear. Nima’s new approach also describes the world in terms of a geometric space, but this space’s meaning is a lot more clear.

The space is called “kinematic space”. That probably still sounds mysterious. “Kinematic” in physics refers to motion. In the beginning of a physics class when you study velocity and acceleration before you’ve introduced a single force, you’re studying kinematics. In particle physics, kinematic refers to the motion of the particles you detect. If you see an electron going up and to the right at a tenth the speed of light, those are its kinematics.

Kinematic space, then, is the space of observations. By saying that his approach is based on ideas in kinematic space, what Nima is saying is that it describes colliding particles not based on what they might be doing before they’re detected, but on mathematics that asks questions only about facts about the particles that can be observed.

(For the experts: this isn’t quite true, because he still needs a concept of loop momenta. He’s getting the actual integrands from his approach, rather than the dual definition he got from the amplituhedron. But he does still have to integrate one way or another.)

Quantum mechanics famously has many interpretations. In my experience, Nima’s favorite interpretation is the one known as “shut up and calculate”. Instead of arguing about the nature of an indeterminately philosophical “real world”, Nima thinks quantum physics is a tool to calculate things people can observe in experiments, and that’s the part we should care about.

From a practical perspective, I agree with him. And I think if you have this perspective, then ultimately, kinematic space is where your theories have to live. Kinematic space is nothing more or less than the space of observations, the space defined by where things land in your detectors, or if you’re a human and not a collider, in your eyes. If you want to strip away all the speculation about the nature of reality, this is all that is left over. Any theory, of any reality, will have to be described in this way. So if you think reality might need a totally new weird theory, it makes sense to approach things like Nima does, and start with the one thing that will always remain: observations.